Julia in Multi-Agent Systems (MSc Thesis)
Zero-Intelligence trader simulations implemented in Julia with George Mason University researchers, benchmarked against C / C++ / Java / Python.
- role
- Author — MSc thesis
- period
- 2024
Problem
Whether Julia is a viable language for multi-agent simulation — fast enough for the numerical core, simple enough to be worth it.
What was built
- Zero-Intelligence trader simulations in Julia, in collaboration with George Mason University researchers.
- Unified implementations in C, C++, Java, and Julia for a fair performance comparison.
Outcome
- Julia ~1.2× slower than C and comparable to C++, while winning on code simplicity.
- Across the language comparison: Python ~10× slower than Java; Java 2–4× slower than Julia.